JoyVoice: Long-Context Conditioning for Anthropomorphic Multi-Speaker Conversational Synthesis

TL;DR

JoyVoice is a novel multi-speaker conversational speech synthesis model that enables boundary-free, long-form, multilingual, and zero-shot voice cloning with superior naturalness and prosody, using a unified end-to-end transformer architecture.

Contribution

Introduces JoyVoice, a unified end-to-end transformer-based model for flexible, boundary-free multi-speaker conversational speech synthesis with a novel MM-Tokenizer and robust data processing.

Findings

Achieves state-of-the-art multilingual generation and zero-shot voice cloning.

Demonstrates superior prosodic continuity and naturalness in long-form speech.

Outperforms existing models on Seed-TTS-Eval and multi-speaker tasks.

Abstract

Large speech generation models are evolving from single-speaker, short sentence synthesis to multi-speaker, long conversation geneartion. Current long-form speech generation models are predominately constrained to dyadic, turn-based interactions. To address this, we introduce JoyVoice, a novel anthropomorphic foundation model designed for flexible, boundary-free synthesis of up to eight speakers. Unlike conventional cascaded systems, JoyVoice employs a unified E2E-Transformer-DiT architecture that utilizes autoregressive hidden representations directly for diffusion inputs, enabling holistic end-to-end optimization. We further propose a MM-Tokenizer operating at a low bitrate of 12.5 Hz, which integrates multitask semantic and MMSE losses to effectively model both semantic and acoustic information. Additionally, the model incorporates robust text front-end processing via large-scale data perturbation. Experiments show that JoyVoice achieves state-of-the-art results in multilingual generation (Chinese, English, Japanese, Korean) and zero-shot voice cloning. JoyVoice achieves top-tier results on both the Seed-TTS-Eval Benchmark and multi-speaker long-form conversational voice cloning tasks, demonstrating superior audio quality and generalization. It achieves significant improvements in prosodic continuity for long-form speech, rhythm richness in multi-speaker conversations, paralinguistic naturalness, besides superior intelligibility. We encourage readers to listen to the demo at https://jea-speech.github.io/JoyVoice